In the analysis of spectroscopic ellipsometry (SE) data acquired on thin film structures, the weighted error function chi(2) plays a central role, allowing one to assess the validity of the optical model used to generate the best-fit simulation. The most difficult steps in implementing chi(2), however, involve the required quantification of the precision, accuracy and reproducibility of the SE measurement. In real-time SE, spectra are acquired continuously vs. time during the fabrication of thin film structures, and one has the advantage of extracting the conventional spectral averages of chi(2) vs. time or the temporal averages vs. photon energy. These two measures of the validity of the optical model, not only provide insights into the fitting procedure, but also allow one to identify otherwise unrecognized errors that can account for unexplained variations in chi(2) (C) 2003 Elsevier B.V. All rights reserved.